Risk-Aware Machine Learning Framework for Fraud Detection in Financial Technology Ecosystems
Journal of Digital Market and Digital Currency · 02.11.2026 · 3 (4) · 315-333
DOI: 10.47738/jdmdc.v3i4.104 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Digital Market and Digital Currency, online erschienen am 02.11.2026. Verfasst von Richard Berlien.
Abstract (Originalsprache)
Financial fraud has become one of the major challenges in modern financial technology ecosystems due to the increasing volume and complexity of digital transactions. Traditional fraud detection approaches often struggle to identify complex and dynamic fraud behavior patterns, particularly in noisy and imbalanced transaction environments. This study proposes a risk-aware machine learning framework for fraud detection in fintech ecosystems by integrating contextual risk indicators, behavioral transaction features, and explainable artificial intelligence techniques. The proposed framework incorporates feature engineering, SMOTE balancing, ensemble-based machine learning algorithms, and SHAP interpretability analysis to improve fraud classification performance and model transparency. The experiments were conducted using Random Forest, XGBoost, and LightGBM models on a synthetic fintech transaction dataset enhanced with realistic noise injection and interaction-based risk features. The experimental results demonstrated that LightGBM achieved the best overall performance with an accuracy of 97.75%, F1-score of 80.69%, and ROC-AUC of 0.8891. Furthermore, the ablation study confirmed that contextual risk-aware features, including device risk score, IP risk score, and risk interaction indicators, significantly improved fraud detection capability. The SHAP analysis also revealed that transaction amount, temporal transaction behavior, and interaction-based risk features substantially influenced fraud prediction outcomes. Overall, the findings indicate that combining contextual risk intelligence, behavioral analytics, and explainable machine learning techniques can significantly improve fraud detection performance and interpretability in modern fintech environments. The proposed framework provides a scalable and transparent approach for intelligent fraud analytics in digital financial systems.
Zitieren
APA
Berlien, R. (2026). Risk-Aware Machine Learning Framework for Fraud Detection in Financial Technology Ecosystems. Journal of Digital Market and Digital Currency, 3(4), 315-333. https://doi.org/10.47738/jdmdc.v3i4.104
Vancouver
Berlien R. Risk-Aware Machine Learning Framework for Fraud Detection in Financial Technology Ecosystems. Journal of Digital Market and Digital Currency. 2026;3(4):315-333. doi:10.47738/jdmdc.v3i4.104
BibTeX
@article{Berlien2026,
author = {Berlien, Richard},
title = {Risk-Aware Machine Learning Framework for Fraud Detection in Financial Technology Ecosystems},
journal = {Journal of Digital Market and Digital Currency},
year = {2026},
volume = {3},
number = {4},
pages = {315--333},
doi = {10.47738/jdmdc.v3i4.104}
}
EUVIMED